Audio Scene Generation on ASG-Bench Text
84.34Event (%)Audio-Oscar
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Audio-OscarLLM=DeepSeek-V4-Flash, Thinking=false2026.06 | 84.34 | 82.92 | 4.06 | 4.12 | 3.93 | |
| Audio-OscarLLM=Qwen-397B-A17B, Thinking=false2026.06 | 83.8 | 81.79 | 3.95 | 4.11 | 3.86 | |
| Audio-OscarLLM=DeepSeek-V4-Flash, Thinking=true2026.06 | 81.94 | 81.48 | 3.49 | 3.74 | 3.36 | |
| Audio-OscarLLM=Qwen-122B-A10B, Thinking=false2026.06 | 79.98 | 79.51 | 4.04 | 4.12 | 3.93 | |
| WavJourneyLLM=DeepSeek-V4-Flash, Thinking=false2026.06 | 72.33 | 76.13 | 3.66 | 3.5 | 3.28 | |
| Any2Speech2026.06 | 51.65 | 41.16 | 3.53 | 2.63 | 2.74 |